As a method of structural optimization, the Taguchi method is applied and programmed on a parametric CAD platform. Geometric parameters of CAD models are directly taken as design variables. Parameter design is made using control factors and noise factors selected from the design variables. Analysis of mean(ANOM) and variance(ANOVA) are done based on simulations by finite elements method integrated in the developed program. A sequence of Taguchi solutions are generated in such a way to reduce the region of uncertainty. To treat inequality constraints, a variable penalty method is utilized. A combination of performance index and constraint functionals can be used as an optimal design formulation. The proposed procedure is programmed into a set of menus on the CAD platform. Several examples are taken to examine performance of the Taguchi approach and illustrate the procedure of the design tool. The method is shown applicable for practical structural design. It is, however, not menat to obtain the exact optimum point. It is a versatile and robust tool in finding near optimum solutions.